Confidential

An AI negotiator for consumers.
Fair value for everyone.

August 2026 · counterwell.ai
Marketplace Local Offer Cars
The founder

From research to product.

Claire ChenFounder & CEO
  • Mathematics
  • 6 published papers · ICML ×3, ICLR ×2, AAAI oral (top 4.7%)
  • LLM Agents · Reinforcement Learning · Game Theory
Research collaborators
MITDavid Simchi‑Levi Member, National Academy of Engineering
CornellThorsten Joachims Vice Provost for AI Strategy
BerkeleySergey Levine Berkeley AI Research (BAIR)
BerkeleyPieter Abbeel Berkeley AI Research (BAIR)
Co‑Director
Founder’s journey →
The problem

The negotiation gap.

Used car · $42,900Illustrative scenario
BuyerIf we close today, would $40,500 work?
SellerThat’s my best price.
$42,200. You can drive it home today.
BuyerOkay, $42,200 it is.
Is that really the best price?The buyer cannot see how much room is left.
SellerEvery day · hundreds of deals
BuyerEvery few years · one deal

Consumers start with less information and experience.

Where we start

From everyday purchases to life’s big deals.

High volume
Marketplace
Secondhand goods
AI shopping assistantFinds that fit the budget
Local
Home services
Multi-seller negotiationSide-by-side quotes
High value
Offer
Compensation
Compensation breakdownSalary negotiation rehearsal
Cars
Vehicle purchases
Multi-dealer negotiationCompare out-the-door costs
The solution

Our strategy model × parallel negotiation

Trained for economic outcomes

Optimize buyer gains, not conversational preferences.

OffersConcessionsTiming
Reinforcement learning · Multi-agent games

Make sellers compete for the order

Seller A
Initial $520
Now$455
Round 3
Seller B
Initial $500
Now$425
Round 4 · Best offer
Seller C
Initial $535
Now$448
Round 2
Best offer$425
Below the lowest initial quote: $75
Illustrative interaction

Consumers set the goal. Counterwell negotiates in parallel.

Full workflow in the appendix →
See it work

One system, four kinds of negotiation.

Watch the films at deck.counterwell.aiWATCH THE FILMS
recorded in the live product · watch at deck.counterwell.ai
Why we win

They optimize for helpfulness. We optimize for outcomes.

General-purpose models

optimized for helpfulness

↓ tends to concede

optimized for buyer outcomes
RLVR

↑ counters strategically

Four LLM negotiation papers · Claire Chen, main author
The research

Four papers. Four forms of negotiation.

One buyer, one seller Instructing LLMs to Negotiate with Verifiable Rewards
One against many Strategic Bargaining in Multi‑Buyer Markets
Many items, one counterparty Multi‑Item Bilateral Negotiation
Many items, many counterparties Counter‑offers that make sellers bid

In all four, the model is trained against opponents that keep getting stronger.

Appendix: Measuring negotiation performance →Learning negotiation strategies →
Claire Chen, main author on all four · Research program in the appendix →
The live test

The smallest model got the best price.

46.27%
Counterwell 30B
our research model
28.35%
GPT-5.4
latest frontier
23.81%
DeepSeek
V3.1 · 671B
19.69%
Humans
our own team
The discount obtained, across 30 real deals on Facebook Marketplace.
sources in the appendix
The moat

Every real negotiation deepens the data moat.

1Capture real negotiations Multiple rounds, multiple opponents
2Link to economic outcomes Price · Deal · Walk-away
3Improve the strategy model Learn which strategies work
4Compound the data advantage More deals → richer data → better strategy

Strategy data labeled with real economic outcomes, not just chat logs.

Data and technical moat in the appendix →

Competition

A strategic agent that fights for the consumer.

Strategic outcome
optimization
General assistance
B2B
B2C
RealPageSalesforceHubSpot
ADPWorkday
AmazonUber
ChatGPTClaudeGemini
Levels.fyiHoney
MuseInstinct
Market potential

Four markets. One platform.

Money in play, every year, US
$200B+
MarketplaceUS secondhand goods
$842B
LocalHome services
$1.8T
Cars54M vehicles, new and used
$13T
OfferWages and salaries
Total$15.84T
How that becomes revenue
$15.84Tnegotiable every year
×
1%negotiated through us
×
~0.1%current pricing equivalent
=
$158Mannual revenue
A subscription, not a commission; the rate is subscription revenue over the value negotiated.
$158M$158B negotiated1% of the market
$317M$317B negotiated2% of the market
$792M$792B negotiated5% of the market
$1.58B$1.58T negotiated10% of the market
How we make money

Flexible subscriptions, by negotiation type.

Everyday
MarketplaceLocal
$4.99 / mo · $9.99 / mo
The big ones
CarsOffer
$39 / mo · $49 / mo
Everything
Counterwell Pro
$59 / mo · $179 / year
Free trial to start negotiating. Monthly plans cancel any time, unlimited inside the month; annual options are available on selected plans.
Backing

$38,000 in, before any dilution.

$20,000
Ryan Summer Program
$10,000
Bill Gross Prize
$5,000
Tinker credits
THINKING MACHINES
$3,000
Demetriades Prize
The plan

From building the product to growing the users.

Built All four products built
Today
What the campaign will answer
User conversion
Seller response rate
Market priority
The ask

$1.3M. Raising now.

Pre-seed, open now
$1.3M
Pre-seed. Eighteen months. The product is built.
Seed round at month twelve, after at least 500 paying customers.
52%People Founder full-time and two engineers
19%First users One campaign at a time, then scale the winner
8%Legal and running costs Incorporation, terms, the car review
3%Office Remote for six months, then San Francisco
2%AI and infrastructure Cheaper than the people by a mile
16%Reserve Held back
The platform view

Making deals happen that otherwise would not.

Fixed pricing

The buyer walks away
Seller cost Buyer budget Fixed price Unrealized gains from trade
No dealGains for both sides: 0

AI agent negotiation

Find a feasible price
Seller cost Buyer budget Negotiated price Seller surplus Buyer surplus
Deal reachedBoth sides gain

Lower negotiation costs. Unlock new deals. Both sides gain.

The long view

When commerce runs on agents,
who represents the consumer?

General models · Agreeableness

Too agreeable. Too quick to concede.

An agent that keeps conceding to reach agreement can give away value it could have secured.

The risk

Strategy gaps become economic gaps.

When one side plays strategically and the other gives in, the economic gap can widen with every deal.

Counterwell’s answer

More value for consumers.

  • Read the other sideInfer motives and limits
  • Act strategicallyTime offers and concessions
  • Use information wiselyTurn needs into better terms
Give consumers a strategy that turns information and needs into better terms.
Why me

I keep choosing the harder path.

1
China → U.S.

Left the sure path

Top provincial school, clear path through Gaokao. Chose the U.S. instead — no SAT, no playbook — and got in.

→
2
UVA

Started behind, rose to the top

Arrived years behind peers. Top grade in graduate ML — ahead of PhD students — and first author at ICLR as a sophomore.

→
3
UVA → Caltech

Restarted at the peak

Left a winning position — grades, papers, mentors — to transfer to Caltech, one of a handful admitted.

→
4
Caltech → Counterwell

Built, not assigned

Originated the research, brought together collaborators, secured compute. That work became Counterwell’s core model.

The founder’s job — done four times before the company existed.

Thank you.

Appendix

The films, the papers, and where every number came from.
Appendix

The four films, in full.

Watch the films at deck.counterwell.aiWATCH THE FILMS
recorded in the live product · watch them at deck.counterwell.ai
Counterwell Marketplace · live

One button, on the page you are already looking at.

The browser extensionSitting beside the listing. One tap reads the page and writes your opening message — nothing to paste.
The full negotiationEvery round after that: what they said, what to send, and how far you are from your ceiling.
Marketplace
Counterwell Local · live

One sentence in. Five companies out.

Say it, or tap a trade. No form.
Names, ratings, review counts, phones, distance — live from Google, with each company’s own photo.
  • Ranked on the evidence. A rating counts for more when more reviews stand behind it, so a 5.0 from 18 does not beat a 5.0 from 76. Google’s own order settles ties and nothing else.
  • Distance can change it — deliberately. Somebody has to drive to you.
Localworking · live Google data
Counterwell Offer · live

Your offer, broken down —
and what to ask for.

Check the numbers$316,200 in year one, $276,200 every year after — because $40,000 of it lands once. The letter never puts those two numbers side by side.
The planLead with the competing offer because its deadline is sooner. Ask for equity, not base — $60,000 to $115,000, settle at $90,000. Words you can say out loud, with no blanks.
Offerbuilt · running today
Counterwell Cars · live

The lowest price is not the cheapest deal.

Kestrel Toyota’s sticker is $996 cheaper — and costs $2,361 more once the rate and the term are counted.

Every offer, cheapest firstEach card carries its gap to the cheapest.
Or drawn, on one rulerEvery charge split into the same four buckets, so a low price with a big fee stops looking like one.
Cars
Appendix · Full workflow

Not a reply. The whole negotiation.

1

Read the situation

Understands the goal from screenshots, files or voice.

2

Find the opening

Finds the quotes and counterparties worth pursuing.

3

Set the strategy

Our own post-trained model decides every move.

4

Negotiate for them

Sends and replies on its own, across sellers and rounds.

5

Rehearse it

Simulates the other side before the call or meeting.

Not a general chatbot — a strategic negotiation system.

Marketplace Local Offer Cars
← Back to the solution
Appendix

Why we beat ChatGPT and Claude.

1 Our model is trained to win. Theirs are trained to agree. General‑purpose models are rewarded for being agreeable, because that is what everyday help, coding and support need. Winning a negotiation rewards the opposite — and that is different model weights, not a different prompt. trained to win
2 Specialization beats brute‑force scale A whole system that does one job — win‑win deals. 46% off against 28% for GPT‑5.4, from a model a tenth the size. backed by research
3 We automate the whole negotiation ChatGPT and Claude hand the user the next sentence to say. We write to every seller ourselves, run all of those conversations at once, and keep pushing for days. full automation
4 Every deal makes the next one better Because we focus on negotiation, every deal ends in a real outcome — a price agreed, or a walk‑away — that is worth training on. Generic conversation never produces one. data flywheel
Counterwell against the general-purpose models
The research program

Research program

1 · Games and learning
How do agents learn
to act strategically?
2published papers
ICML ×2

Markov games
Two-timescale convergence

3 under review
Multi-agent games
2 · Information and evaluation
How can limited data
evaluate a policy?
3published papers
ICLR ×2 · AAAI Oral ×1

Safety-constrained · Multi-policy
Doubly optimal policy evaluation

3 under review
Policy evaluation · Information games
3 · Negotiation strategies
How can buyers
secure better outcomes?
4papers under review
One-to-one · One-to-many

Multiple items and opponents
Across market structures

Reward buyer gains
Train negotiation strategies directly
Foundations
ICML ×1
Formal mathematics · Verifiable reasoning
Real-time decisions at scale Under review
Optimization in high-dimensional spaces
Full publication list in the CV
Appendix

Related research, and where it stands

1 · Games
Offline Two‑Player Zero‑Sum Markov Games with KL Regularization
ICML 2026 · published
Pessimism‑Free Offline Learning in General‑Sum Games via KL Regularization
NeurIPS · under review
Fast Rates in α‑Potential Games via Regularized Mirror Descent
NeurIPS · under review
Beyond Pessimism: Offline Learning in KL‑regularized Games
NeurIPS · under review
2 · Information
Pessimistic Minimax Learning for Public‑Private Information Games under Unilateral Coverage
NeurIPS · under review
Breaking the Curse of History in LQG: Polynomial Off‑Policy Evaluation via Marginalized Importance Sampling
Working paper
Online learning as private information becomes public: does the policy converge?
Direction
3 · Negotiation
Instructing LLMs to Negotiate using Reinforcement Learning with Verifiable Rewards
arXiv 2604.09855
Strategic Bargaining in Multi‑Buyer Markets
arXiv 2607.05863
Learning to Recommend: Multi‑Item Bilateral Negotiation via LLM Sellers
Under submission
Learning to Sell: Multi‑Product Portfolio Allocation via LLM Agents
Under submission
At scale
OR‑Transformer: Scaling Real‑Time Decision‑Making to 1,000 Items
arXiv 2609.01933 · under review
Strategic Discovery in Agentic Commerce
Direction
Buyer‑Private Mechanism Design for Agentic Commerce
Direction
Claire Chen leads all of the above.
Full publication list in the CV
Appendix · Measurement

Measuring negotiation performance

The negotiating room, as a tug of war
Buyer bargained ratio
← Back to the research overview
Appendix · Training

Learning negotiation strategies

Reward Buyer bargained ratio First-offer / budget ratio Deal ratio
← Back to the research overview
Appendix

82% joined the waitlist on the spot.

survey · 60 Caltech students
Appendix

Where those numbers come from.

FigureWhat it isSource
$13.0TUS wages and salaries paid, 2025 Bureau of Economic Analysis, national accounts
$842BUS home services market, 2026 Mordor Intelligence — US home services market definition
$1.8TUS vehicles sold in 2025: 16.3M new × $50,000, plus 37.8M used × $25,730 Cox Automotive for the volumes; Kelley Blue Book for both prices
$200B+US secondhand goods bought in a year — furniture, electronics, tools, sporting goods, clothes. Vehicles are not in it, so the row above is not counted twice. $306.5B projected for 2030 OfferUp / GlobalData, 2025 Recommerce Report
$50Expected revenue per acquired paying customer, blended across the four products and their expected tenures Our own prices and how long each kind of customer stays — worked out on the next slide
134.8MUS households, 2025 — what the percentages under the bars are a share of US Census Bureau
82%49 of 60 students joined the waitlist on the spot — interest, not paid signups Our own survey, asked in person on one campus
$50 is the modeled expected revenue per acquired paying customer, blended across the current plans and their expected tenures. The cash-flow projection is cohort-based: a customer contributes receipts only while they are active. Counterwell takes no percentage of transaction value, so the market figures above do not directly determine our revenue.
sources checked 12 August 2026
Appendix

What a customer pays, and what the AI costs to serve them.

Expected revenue per acquired paying customer
They came forPrice × modeled tenureOf 100Together
Marketplace$4.99 / mo × 8-month modeled tenure = $4050$2,000
Local$9.99 / mo × 2.5 = $2515$375
Cars$39 / mo × 1 month = $3915$585
Offer$49 / mo × 1 month = $4912$588
Counterwell Pro$179 for the year8$1,432
100 paying customers100$4,980
$4,980 ÷ 100 ≈ $50 each. Cars and Offer are bought for one month — that is how long the job takes — and Marketplace is modeled at eight. Sensitivity: if Marketplace customers instead take the $49 annual plan, the blend is about $54.
AI inference cost — per acquired customer, over the same tenure
What they use over that tenureAI costAI-level margin
About 36 negotiations$2.3894.1%
About 6 jobs$1.0795.7%
6 cars looked at, 1 bought$2.3194.1%
3 offers, 6 calls rehearsed$1.8296.3%
All four, heavily$5.1897.1%
Moonshot’s published rates: $0.95 per million tokens in, $4.00 out. A whole car negotiation is 39¢. AI cost applies a full year’s usage allowance over the shorter modeled tenure, which is deliberately conservative.
$50 in, $2.33 of AI cost out — a 95.3% margin at the AI level. Each product serves a different customer need. The $50 is a blended planning assumption across customer types and expected tenure, not an assumption that one person buys several products.
prices checked 14 August 2026
Appendix

Where the $1.3M goes.

AmountWhat it pays forOf the round
rounded
$210,000Founder at $140,000 a year for all eighteen months — within the $132–153k seed-stage band16%
$219,000A founding engineer on $175,000 a year from month four. Median base is $195,000; $175,000 is the seed rate, trading cash for equity17%
$120,000A second engineer on $160,000 a year from month ten, once there are users to serve9%
$121,000Payroll tax, benefits and insurance, at 22% of salary9%
$250,000Marketing — paid customer acquisition, modeled at $4 per signup. $4/signup and 8% signup-to-paid conversion are pre-pilot assumptions; actual campaign results will replace both19%
$100,000Incorporation, consumer terms and privacy, the automotive review, accounting and tools8%
$30,000AI and infrastructure — model calls behind every negotiation, quote and letter, plus operating headroom, for all eighteen months2%
$42,000A shared office in San Francisco from month seven, at $3,500 a month — remote until then3%
$208,000Runway and contingency reserve16%
$1.3M = $1,092,000 of planned spend + $208,000 held as reserve. Under the current model the raise acquires roughly 5,000 paying customers over eighteen months, projecting about $216,000 of customer cash receipts. The model assumes six months remote, then twelve in a shared San Francisco office. The next page shows the cash-flow timing.
Appendix

18-month cash flow projection.

Cumulative
customers acquired		Active paying
at period end		 Operating spendCustomer
cash receipts		Net cash burn
Months 1–6750520$235,000$28,000$207,000
Months 7–122,2501,240$377,000$66,000$311,000
Months 13–185,0002,300$480,000$122,000$358,000
Month 18 / 18-month total5,0002,300 $1,092,000$216,000$876,000
Spending is budgeted; customer receipts are projected. Receipts are cohort-based: a customer pays only during the modeled tenure for their product, so cumulative customers acquired and active paying customers differ. Marketing is phased 15% / 30% / 55% across the three periods, as the channel moves from validation to scaling. Some receipts from customers acquired late in the runway fall beyond month eighteen.

Acquisition is modeled at $4/signup and 8% signup-to-paid conversion, both pre-pilot assumptions — together they imply a $50 cost per paying customer acquired. Month-18 customer cash receipts are approximately $21,800.
Appendix

What month twelve looks like.

Milestone one

At least 500 paying customers

Roughly $2,500 a month, recurring.

  • 1,240is what the model on the previous page projects
  • 500is the number we plan to hold
Milestone two

One channel that repeats

All four products tested, then the winner scaled.

  • 45%of the marketing budget spent by month twelve, finding the channel
  • 55%held back for months thirteen to eighteen, to scale what worked

The model’s customer numbers follow from $4 a signup and 8% signup‑to‑paid, both pre‑pilot assumptions the first campaign replaces — so the milestone is set well under them.

the plan, not results
00:00

Say

Cues

Then

Next

Use → ← to move · F fullscreen